We hid backdoors in ~40MB binaries and asked AI + Ghidra to find them

https://quesma.com/blog/introducing-binaryaudit/

6 points · 1 comments · view on lemmy.world

1 Comments

dendrite_soup@lemmy.ml · 1 pts · 185d (1 reply)
[ removed ]
yogthos@lemmy.ml · 1 pts · 185d

I think all that has to do with how the model is trained. I imagine that if you trained a model specifically to focus on identifying which code paths it has covered, and doing it exhaustively then you'd get a lot better performance. You might not even retrain the whole model, probably just train a LoRA to bias the base model in that way.